Updated files for 0.2.73 release

  * Fixed a nonsense in configure.in preventing to build sdl.so.
  * Removed a few stupid debug instructions in aout_s16.c that should
    have never been commited. Sorry about that.

  * Don't use the #@%&$! autoconf-suggested CFLAGS.

  * Coding style fixes here and there.
  * Miscellaneous QNX compile fixes.
  * Beginning of the mingw32 port by Gildas Bazin <gbazin@netcourrier.com>.
  * Added Marcari and Gildas to the AUTHORS file.

  (I commited the mingw32 port because it doesn't add any new code outside
   the #ifdefs, but we won't tell a word about it for the moment. We don't
   really need windows users whining about bugs, do we ?)

Added a howto. Feel free to correct it. I'll put it on the website soon.

done inside a #ifdef DEBUG

It still segfaults for me on exit.

--enable-debug still exists, but only activates -g and additional
safety checks.

I hope I didn't break anything :))

Repaired broadcast support : binding INADDR_ANY doesn't seem to work.

You now have to specify the broadcast address as an argument.
Btw, thanks to sam, you can now use ts://server:port style, which gives
for broadcast :

vlc --broadcast broadcast_addr ts://server[:port]

  * Fixed the continuous 'seeking position' bug in network mode.
  * rc interface plugin for vlc control through /dev/stdin courtesy
    of Peter Surda <shurdeek@panorama.sth.ac.at>.
  * Support for `ts://foo:42' style input source:
     vlc ts://vls.via.ecp.fr:1234
     vlc ts://vls.via.ecp.fr
     vlc ts:vls
     vlc ts:vls:4321
    As a side effect, the Gnome and Gtk "network input" buttons work.

  * Fixed the ac3dec compilation under BeOS.
  * Removed a useless #include in aout_sdl.c.

  * Commited Shane's XVideo patches. It works here (tm).
  * Fixed the XVideo aspect ratio (AR_SQUARE_PICTURE means square pixels,
    not square image).

  - New ac3_mantissa (use of dithflag, new unpack)
  - Correction of an introducing counting bug in ac3_parse
    (i forget 2 bits in total_bits_read)

callback on startup - InitBitstream prototype has changed ;
* Removed a buffer zeroing in video output - doesn't change anything
for me, does it for you ?

  * Updated the TODO list.
  * Fixed benny's latest C++ism.
  * Fonts not found are no longer fatal to the video output.

check if the new code still compiles everywhere, but it should.

-removed � bug added at my last commit in css code; cleaned CSSGetKey

-various changes and bugfixes in the gnome interface.

  * XVideo fullscreen mode by David Kennedy <dkennedy@tinytoad.com>.

  * X11 fullscreen mode, courtesy of David Kennedy <dkennedy@tinytoad.com>

  * Added prototypes for the Elementary Stream input module for Alexis.
     it's compiled by default, to activate it: "vlc --input es <file.mpeg>"

  - New ac3_decoder_thread (we now use GetBits)

Please warn me if you encounter some problem

  * Fixed build process when no plugins or no builtins were chosen.
  * Makefile changes to get around MacOS X's brain-damaged gcc.
  * FreeBSD configure and Makefile patches, courtesy of Espen
    Skoglund <esk@ira.uka.de>.

synchronized ; and menus have less chances to be changed when they're opened.

-added basic angle support ; this is a first try. It works with my 2
test-DVDs, but I'd like that people having DVDs with repeated scenes to
test it. We can choose angle with command line option -u (don't ask me
why :)

-a cosmetic change in css decryption.

Since config.h.in has changed, you have to re-run ./configure.

<sbrown@cortland.com>)

   OS X fixes:
  * Fixed a compilation issue in the DVDioctl driver.
  * Added a call to ranlib after a built-in module compilation.

  * Fixed a typo in configure.in

  * Merged Makefile.common and Makefile.modules.in into Makefile.modules
  * Updated version number and ChangeLog to 0.2.72
  * Additional plugin Makefile cleaning, probably the last.

Re-added the new packet allocation method in PS input, using packet caches.
With warning level 1, you can see when a packet is allocated or freed; the
goal is to never call 'malloc' nor 'free', except at the beginning of the
stream.
The size of the different caches can be adjusted in plugins/mpeg/input_ps.h.
If 'free' is called too often, it means that a cache is too small, so try
to increase its size.

  * Added .cvsignore files in the plugins directories.
  * Additional Makefile cleaning for built-in modules.
  * Fixed Makefile for Solaris target.

have a dvd drive here, but it should work now.

  * All Debian packages are now lintian-clean.
  * Fixed XVideo plugin linking.

  * Updated TODO list with Henri's Apr 6th post.
  * Built-in modules support almost finished.

    Now you can compile any module within vlc, even the Gnome, Qt, SDL
  and GGI modules at the same time, resulting in an awful extra fat
  vlc binary with gazillons of library dependecies.

    The only exception is that the Gnome and Gtk+ modules can't be
  compiled together into vlc. But you can have one compiled built-in
  and the other as a plugin.

    Another consequence is that libdl should no longer be needed under
  MacOS X. I'll check this as soon as possible.

  * A few Makefile fixes for BeOS
  * XVideo patches by Shane Harper

  * Beginning of the built-in modules support.

   A few words about the changes:

  - heavy Makefile butchery has taken place, each plugin now has its own
    Makefile. I know recursive make blablah harmful, but it was just so
    much easier to do this way.

  - Makefile.in has disappeared, we now generate Makefile.opts with
    the overall configuration options, and Makefile.modules which is
    specific to module compilation.

  - After ./configure has been run you may want to modify Makefile.opts
    to check which modules will be compiled built-in or as plugins.

  - Some modules cannot be compiled built-in right now because proper
    linkage doesn't work yet. We don't really care since they're the
    interface or video output modules. The most important stuff works
    (iDCT, motion, YUV, input).

  - It's perfectly valid to compile a module both as built-in and as a
    plugin. vlc will only load the built-in one, but I'll add an option
    to ignore built-in modules for testing purposes.

  - We *should* see a performance increase here. I didn't have much time
    to test it, but if anyone can confirm and perhaps give a rough
    estimate of how much we gain...

  * Got rid of the sleep() hack in beos_specific.cpp
  * Fixed a segfault when launched without argument under Darwin.